The engineering team wanted to show users every step the protocol was taking. My hunch, competitive research, and observed user testing all disagreed. I convinced the technical founder to automate the final confirmation step — changing the underlying tech, not just the UI. Mental model shifted from 'what is the machine doing' to 'I sent BTC, I'm receiving WBTC.' Flow cut in half. Completion rate up 60%.

Our brand became a trust signal. Investors cited it unprompted. The system I built works so well it runs without me.

In a space built on hype and anonymity, we went the other way. The insight came from a simple Christmas video — real people, real hands — that resonated deeply with the investors and partners we wanted to reach. We built around that: restraint, materiality, negative space over digital spectacle. One system covering identity, motion, social graphics, and tone of voice. Authentic, confident, built to last.

A token-synced, AI-native pipeline that's actively changing how the team ships.

Tokens sync directly from Figma to the codebase. Claude Code reads design intent via MCP, queries existing Storybook components, and writes production-ready output rather than specs that still need to be built. When a component doesn't exist yet, it gets flagged automatically, built in Storybook in isolation, and promoted back into the layout — so the system grows from real product work rather than upfront abstraction.

The result: a three-phase pipeline where designers retain judgment on IA and flows, the design system handles consistency mechanically, and developers ship with real components and real data.

One designer, two engineers, one founder. Swap flow, lending flow, marketing site, blog, and a social system — designed and shipped in 10 days.

The constraint was the point. No room for ceremony, no extended review cycles. Every decision had to be defensible and shippable the same day. Patterns from previous large-scale product work moved fast without sacrificing quality.

Some projects are about shipping. This one is about thinking. Lightspeed is a self-initiated design exploration — no client, no deadline pressure, just a problem I found worth solving and a process made fully visible. Every decision is documented: what I saw, what I considered, what I cut and why. If the other sections show what I ship, this one shows how I get there.